/* * Copyright 2010-2015 JetBrains s.r.o. * * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. * You may obtain a copy of the License at * * http://www.apache.org/licenses/LICENSE-2.0 * * Unless required by applicable law or agreed to in writing, software * distributed under the License is distributed on an "AS IS" BASIS, * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. * See the License for the specific language governing permissions and * limitations under the License. */ package kotlin.reflect.jvm.internal.pcollections; /** * A non-public utility class for persistent balanced tree maps with integer keys. * <p/> * To allow for efficiently increasing all keys above a certain value or decreasing * all keys below a certain value, the keys values are stored relative to their parent. * This makes this map a good backing for fast insertion and removal of indices in a * vector. * <p/> * This implementation is thread-safe except for its iterators. * <p/> * Other than that, this tree is based on the Glasgow Haskell Compiler's Data.Map implementation, * which in turn is based on "size balanced binary trees" as described by: * <p/> * Stephen Adams, "Efficient sets: a balancing act", * Journal of Functional Programming 3(4):553-562, October 1993, * http://www.swiss.ai.mit.edu/~adams/BB/. * <p/> * J. Nievergelt and E.M. Reingold, "Binary search trees of bounded balance", * SIAM journal of computing 2(1), March 1973. * * @author harold */ final class IntTree<V> { // marker value: static final IntTree<Object> EMPTYNODE = new IntTree<Object>(); // we use longs so relative keys can express all ints // (e.g. if this has key -10 and right has 'absolute' key MAXINT, // then its relative key is MAXINT+10 which overflows) // there might be some way to deal with this based on left-verse-right logic, // but that sounds like a mess. private final long key; private final V value; // null value means this is empty node private final IntTree<V> left, right; private final int size; private IntTree() { size = 0; key = 0; value = null; left = null; right = null; } private IntTree(long key, V value, IntTree<V> left, IntTree<V> right) { this.key = key; this.value = value; this.left = left; this.right = right; size = 1 + left.size + right.size; } private IntTree<V> withKey(long newKey) { if (size == 0 || newKey == key) return this; return new IntTree<V>(newKey, value, left, right); } boolean containsKey(long key) { if (size == 0) return false; if (key < this.key) return left.containsKey(key - this.key); if (key > this.key) return right.containsKey(key - this.key); // otherwise key==this.key: return true; } V get(long key) { if (size == 0) return null; if (key < this.key) return left.get(key - this.key); if (key > this.key) return right.get(key - this.key); // otherwise key==this.key: return value; } IntTree<V> plus(long key, V value) { if (size == 0) return new IntTree<V>(key, value, this, this); if (key < this.key) return rebalanced(left.plus(key - this.key, value), right); if (key > this.key) return rebalanced(left, right.plus(key - this.key, value)); // otherwise key==this.key, so we simply replace this, with no effect on balance: if (value == this.value) return this; return new IntTree<V>(key, value, left, right); } IntTree<V> minus(long key) { if (size == 0) return this; if (key < this.key) return rebalanced(left.minus(key - this.key), right); if (key > this.key) return rebalanced(left, right.minus(key - this.key)); // otherwise key==this.key, so we are killing this node: if (left.size == 0) // we can just become right node // make key 'absolute': return right.withKey(right.key + this.key); if (right.size == 0) // we can just become left node return left.withKey(left.key + this.key); // otherwise replace this with the next key (i.e. the smallest key to the right): // TODO have minNode() instead of minKey to avoid having to call get() // TODO get node from larger subtree, i.e. if left.size>right.size use left.maxNode() // TODO have faster minusMin() instead of just using minus() long newKey = right.minKey() + this.key; //(right.minKey() is relative to this; adding this.key makes it 'absolute' // where 'absolute' really means relative to the parent of this) V newValue = right.get(newKey - this.key); // now that we've got the new stuff, take it out of the right subtree: IntTree<V> newRight = right.minus(newKey - this.key); // lastly, make the subtree keys relative to newKey (currently they are relative to this.key): newRight = newRight.withKey((newRight.key + this.key) - newKey); // left is definitely not empty: IntTree<V> newLeft = left.withKey((left.key + this.key) - newKey); return rebalanced(newKey, newValue, newLeft, newRight); } /** * Changes every key k>=key to k+delta. * <p/> * This method will create an _invalid_ tree if delta<0 * and the distance between the smallest k>=key in this * and the largest j<key in this is |delta| or less. * <p/> * In other words, this method must not result in any change * in the order of the keys in this, since the tree structure is * not being changed at all. */ IntTree<V> changeKeysAbove(long key, int delta) { if (size == 0 || delta == 0) return this; if (this.key >= key) // adding delta to this.key changes the keys of _all_ children of this, // so we now need to un-change the children of this smaller than key, // all of which are to the left. note that we still use the 'old' relative key...: return new IntTree<V>(this.key + delta, value, left.changeKeysBelow(key - this.key, -delta), right); // otherwise, doesn't apply yet, look to the right: IntTree<V> newRight = right.changeKeysAbove(key - this.key, delta); if (newRight == right) return this; return new IntTree<V>(this.key, value, left, newRight); } /** * Changes every key k<key to k+delta. * <p/> * This method will create an _invalid_ tree if delta>0 * and the distance between the largest k<key in this * and the smallest j>=key in this is delta or less. * <p/> * In other words, this method must not result in any overlap or change * in the order of the keys in this, since the tree _structure_ is * not being changed at all. */ IntTree<V> changeKeysBelow(long key, int delta) { if (size == 0 || delta == 0) return this; if (this.key < key) // adding delta to this.key changes the keys of _all_ children of this, // so we now need to un-change the children of this larger than key, // all of which are to the right. note that we still use the 'old' relative key...: return new IntTree<V>(this.key + delta, value, left, right.changeKeysAbove(key - this.key, -delta)); // otherwise, doesn't apply yet, look to the left: IntTree<V> newLeft = left.changeKeysBelow(key - this.key, delta); if (newLeft == left) return this; return new IntTree<V>(this.key, value, newLeft, right); } // min key in this: private long minKey() { if (left.size == 0) return key; // make key 'absolute' (i.e. relative to the parent of this): return left.minKey() + this.key; } private IntTree<V> rebalanced(IntTree<V> newLeft, IntTree<V> newRight) { if (newLeft == left && newRight == right) return this; // already balanced return rebalanced(key, value, newLeft, newRight); } private static final int OMEGA = 5; private static final int ALPHA = 2; // rebalance a tree that is off-balance by at most 1: private static <V> IntTree<V> rebalanced(long key, V value, IntTree<V> left, IntTree<V> right) { if (left.size + right.size > 1) { if (left.size >= OMEGA * right.size) { // rotate to the right IntTree<V> ll = left.left, lr = left.right; if (lr.size < ALPHA * ll.size) // single rotation return new IntTree<V>(left.key + key, left.value, ll, new IntTree<V>(-left.key, value, lr.withKey(lr.key + left.key), right)); else { // double rotation: IntTree<V> lrl = lr.left, lrr = lr.right; return new IntTree<V>(lr.key + left.key + key, lr.value, new IntTree<V>(-lr.key, left.value, ll, lrl.withKey(lrl.key + lr.key)), new IntTree<V>(-left.key - lr.key, value, lrr.withKey(lrr.key + lr.key + left.key), right)); } } else if (right.size >= OMEGA * left.size) { // rotate to the left IntTree<V> rl = right.left, rr = right.right; if (rl.size < ALPHA * rr.size) // single rotation return new IntTree<V>(right.key + key, right.value, new IntTree<V>(-right.key, value, left, rl.withKey(rl.key + right.key)), rr); else { // double rotation: IntTree<V> rll = rl.left, rlr = rl.right; return new IntTree<V>(rl.key + right.key + key, rl.value, new IntTree<V>(-right.key - rl.key, value, left, rll.withKey(rll.key + rl.key + right.key)), new IntTree<V>(-rl.key, right.value, rlr.withKey(rlr.key + rl.key), rr)); } } } // otherwise already balanced enough: return new IntTree<V>(key, value, left, right); } }